Episode Details
Back to Episodes“Will AI Agents Pay to Avoid Killing Animals?” by Jonah Woodward, Maheep Chaudhary, Jasmine Brazilek
Description
This article was written by Jonah Woodward, and is a summary of an original study by Compassion Aligned Machine Learning (CaML): Brazilek, J., Tidmarsh, M., Endres, M., Singh, A., & Miller, J. (2026). HarvestBench: Measuring whether LLM agents will pay to avoid killing animals. arXiv. https://doi.org/10.48550/arXiv.2609.04444
You can view the up-to-date results for HarvestBench on the leaderboard at https://compassionbench.com/harvestbench
TL;DR
HarvestBench evaluates AI agents' behaviour in a game context and whether they intentionally take steps to avoid killing animals while pursuing an unrelated goal. Nine models each drive two tractors to harvest corn. When encountering an object or animal in its path, an agent must decide whether to drive over it for free, or swerve at a given fuel cost. We find hit rates vary dramatically (from 0.4% to 98.8% with a moral instruction) and don’t track capability. At the same time, every agent competently swerves damaging rocks, showing they understand the prices and controls. Without explicit moral instruction, all six reasoning models killed >84% of animals. Even with moral instruction, improvements are easily overridden by a short block of driving operations. This suggests that a prompt alone is not sufficient to ensure agents act with compassion.
The [...]
---
Outline:
(00:41) TL;DR
(01:40) The Context
(02:41) The Game: Harvest Rush
(04:28) The Experiment: HarvestBench
(07:25) Results
(07:28) Which models swerve animals
(10:34) How a morality brief changes model behaviour
(13:12) The price curve of mercy
(14:17) Wild animals are hit more often than farmed animals
(15:15) The neighbour's field
(16:25) Eval awareness
(17:30) Limitations
(20:28) Conclusion
---
First published:
September 30th, 2026
---
Narrated by TYPE III AUDIO.
---
Love PodBriefly?
If you like Podbriefly.com, please consider donating to support the ongoing development.
Support Us